Construction and Experimental Evaluation of Document Causality Extraction Model Based on CGCN-BERT

With the increasing degree of networking, the amount of document data also increases. The document causality extraction method is a new data regression analysis technology. In this paper, the Central Graph Convolutional Networks-Bidirectional Encoder Representation from Transformers (CGCN-BERT) text classifier was used to label and filter keywords. This approach can make it easy to extract and judge the causal relationship of documents. Therefore, this paper explored the construction method of document extraction model under this model. This paper mainly used the methods of statistical analysis and experimental comparison to study the document relationship extraction model of CGCN-BERT. The experimental data showed that in the three document tests, the average recall rate was 86.57%, but its accuracy was as low as 67.77%, and the relationship value G was 78.63%. Therefore, the relationship extraction model designed in this paper has some room for improvement.